Use of similarity indices to evaluate responses of vegetation and wildlife to restoration activities

Assessment of the effectiveness of ecological restoration is a difficult problem because temporal monitoring data are usually limited by cost. Probabilistic approaches for assessing community change may offset the difficulties of sparse data sets. In particular, new unbiased estimators of compositional similarity may be useful because they incorporate both relative abundance and unseen shared species. We applied this approach to monitoring data for a dam mitigation project in Washington and Idaho. We chose 12 sites to represent reference conditions for 5 habitats (deciduous forest, floodplain meadow, cattail marsh, shrub, and sedge) and 21 treatment sites, matched to reference sites, to monitor effects of restoration activities (e.g., plantings and weed control). Vegetation and terrestrial vertebrates were surveyed at reference sites for 3 years to evaluate annual variation in community composition. Treatment sites were monitored at 3-year intervals. Avian survey data from 2002-2006 illustrate an issue resolved by the new estimators. From a total of 125 bird species, 26 were observed <10 times, whereas the most common species was observed 1,244 times. For bird species, interyear similarities for reference sites averaged 84% (SD = 8%), whereas similarity of treatment sites to matched reference sites averaged 63% (SD = 13%) with the probabilistic Chao-Jaccard index. Classic Jaccard estimates based on presence/absence were typically 25-30% lower. Our combined data for both vegetation and vertebrates suggest that significant temporal changes due to restoration may be detected by comparing compositional similarity of treated and reference sites using probabilistic estimators.